1. Yva.ai knowledge base
  2. Deep dive
  3. Passive analytics. Employee burnout

Employee Burnout. Yva.ai methodology

Manifestations of burnout depend on its stage. Judging from an employee's appearance, it is often impossible to determine exactly whether they burn out or not. There are tests for self-assessment of stress and burnout, but this solution has serious drawbacks.

For example, how could I be sure that I am experiencing chronic stress if I do not quite understand which level of stress is acceptable and which is not? The answers will vary greatly depending on personality of the employee: someone under the heaviest stress will consider that everything is okay, and someone else, whilst there is a slightest discomfort, will assess their condition as critical.

The test hardly can be scaled to the entire company as the results would be too subjective. Moreover, it would be difficult for employees to pass it regularly.

To solve the problem of burnout assessment, Yva.ai relies on more unbiased passive analytics, i.e. on analysis of employee behavior in corporate communication channels such as email and corporate messengers.

Yva.ai neural network measures both in real time and retrospectively two important indicators: Activity Index (shown as the blue line in the diagram) and Burnout Index (shown as the orange line in the diagram).


Activity Index reflects current state of an employee, their "optimism" or "pessimism"/"frustration", and is formed on the basis of "physical" changes in the digital footprint. Scalar model for calculation of this index is based on 40*N parameters, where N is the number of sources. Linear regression examines five periods using 8 independent variables: workday commencement time, workday length, work visibility, response time, etc.

Burnout Index shows accumulated “optimism” or “pessimism” of an employee. The model is the result of operation of convolutional neural network built on the same 40*N parameters, or rather, on their 40N-dimensional time vectors. Convolution is done so as to maximize the F-measure of resignation prediction. Burnout Index is one of the most significant predictors of possible resigning of an employee. It helps in identification of employees at high risk of resignation 2-6 months after they enter the "red zone".

By default, Yva.ai calculates Activity and Burnout Indices based on metadata of email and other connected sources, without "reviewing" the content of messages. That's why Yva.ai analytics is confidential, ethical and secure.

The system can determine current state of an employee by analyzing relevant historical data. If there is no access to historical data, for example, a newly onboarded employee of the company, then the system will take some time to understand how the employee is behaving properly. The diagram will show you whether the system has enough data to build an accurate model of burnout.

How are Activity and Burnout Indices calculated?

The system builds a standard model of human behavior based on his digital footprint over the past 8 months, and then analyzes how their behavior changes. This analysis is a result of complex machine learning algorithm that takes into consideration 54 metrics, each with its own weight. The neural network calculates the weight of each metric in two combined functions: Activity Index and Burnout Index.

Each parameter is adjusted and individualized. For example, workday length for "night owls" is calculated correctly, even if they finished work on the next calendar day.

The model takes into consideration calendar and production seasonal patterns by week and season. This makes Yva.ai system insusceptible to such seasonal variations in communication as increased activity before the New Year holidays, drop in the number of messages during the official holidays. Seasonal patterns which are unique for the industry/company are taken into consideration only after additional training of the neural network for specific client.

Activity Index reflects changes in the employee's work activity, which is measured across all sources connected to the system. In email as well as in corporate messengers connected to the system, Yva.ai perceives activity of an employee and changes in the activity patterns.

For example, if an employee who used to respond to messages within an hour began to put off their reply and even to ignore some messages, then Yva.ai will understand that they are not behaving as usual, and it is possible that these changes are caused by stress. When the employee becomes more active in correspondence, the blue line ascends, and when they become less active, the line descends.

Decreased activity alone should not be perceived as a warning sign, while even the most active employees need timeouts every now and then. But if the activity descends steadily over an extended period of time, it gives cause for concern. The employee is facing a heavy stress load, and if the cause is not found, the stress can develop into burnout.

Burnout Index shows accumulated “optimism” or “pessimism” of an employee. Burnout Index begins to decrease only if the employee shows chronic slowdown in activity. In such cases, the index becomes negative, and the person moves to the "yellow zone" of burnout. If there is no improvement, and the level of stress does not decrease, then the employee after a while will move to the late stage of burnout or to the “red zone”. The longer a person remains in the red zone, the higher the degree of their burnout, the greater the likelihood of their resignation, the lower their chances of survival in the organization with a 6 month horizon.

Burnout Index is saturated both above and below zero. It means that employees do not accumulate "optimism" or "pessimism" endlessly. The model is trained to show the orange line as reflection of the employee's openness to new offers about 2-6 months before their resignation.

Yva.ai will notify the manager few months before potential resignation of a subordinate, when the employee just begins to feel frustration or burnout. Thus, the company will have a chance to resolve the problem quickly.

Despite positive correlation between burnout and resignation, actual decision to quit depends on many factors, such as family situation, home loan, employment opportunities, etc.

The model ignores data gaps of 2-3 weeks. Vacation will be reflected as a minimal jump of the blue line, the red line will be much about the same.

Transition from one position to another can be reflected in the diagrams. Shortly after such transition (if the role involves less activity) Burnout Index may become negative, then the system will perform recalibration for the employee comparing their current indicators with past indicators, and Burnout Index will return to normal.

How to read the burnout diagram?

This is a burnout diagram of one of the company's employees.


You can see two curves in the diagram.

Activity Index, the blue line, shows "optimism" or "pessimism"/"frustration" of the employee. This index takes into account behavior of the employee in all sources connected to the system, and is calculated on the basis of 40 parameters: workday width, workday commencement time, time of response to messages, intensity of messaging, to mention but a few. When the employee becomes more active in correspondence, the blue line ascends, and when they become less active, the line descends.

Burnout Index, the orange line, shows accumulated “optimism” or “pessimism” of the employee. The Burnout Index begins to decrease only if the employee shows chronic downfall of activity. In such cases, the index becomes negative and indicates that the employee will leave the company in 2-6 months if an opportunity presents itself.

Diagram Fill reflects duration of certain stage of the employee's burnout. Stage of burnout is designated by one of the three color zones.

  • "Green zone" — little or no burnout.
  • "Yellow zone" — frustration (decreased activity) or early burnout.
  • "Red zone" — late burnout, the employee may be already searching for a job.

Width of the zone denotes duration of the employee's permanence at the burnout stage.

Burnout Index is a numerical coefficient, so it can take on the following values.

  • In the "green zone" - positive values only. If an employee is in the “green zone”, the closer their burnout index is to 100, the further the employee is from the burnout risk. The better they are doing.
  • In the “yellow” and “red" zones the values are negative: the closer to -100, the stronger the employee burnout is.

Yva.ai identifies 4 stages before resignation, and is able to detect the first signs of the employee dissatisfaction 4-8 months before they decide to leave the company.

  1. Frustration. Early signs indicating that the employee is dissatisfied with their position or job. The stress level increases. Duration is 4 to 8 months. The employee may continue in the "green area", but their Activity Index diagram takes a tumble.
  2. Early burnout. Passivity of the employee. The employee is in the "yellow zone". Duration is 2 to 4 months. The number of communications decreased, the employee became open to new job offers.
  3. Late burnout. Accumulated pessimism, job search. The employee is in the "red zone". Duration is 1 to 3 months.
  4. Resignation. Handing over the duties. Duration is 2 to 6 weeks. Communication activity may increase during this period.